Online operation of an Artificial NeuralNetwork with adaptive learning rate for speed regulation of DC motor drives

Abstract

DC motor drives are utilized in many industrial plants, including steel 'NOrks, paper factories, mining, conveyers, water pumping and robotic applications. A high dynamic response is highly recommended in industrial applications. Different types of classical controllers are applied on de motors to enhance their dynamic performance. Also, the modern control is applied for many purposes. In this research 'NOrk, novel adaptive Artificial Intelligence (AI)-based digital controllers for de motor speed regulation are developed, simulated and tested in real time through lab proto-type.s. An application of a novel on-line artificial neural net'NOrk (ANN) with an adaptive learning rate is presented. The proposed controller ensures a fast dynamic response for different excursion load conditions. For current protection, a simple rule is introduced in the algorithm to switch on/off the back propagation of the net'NOrk. The adaptive learning rate (Lr) has a nature of PI controller with another expert rule. The proportionality action of L,is driven by the speed error, while the expert rule is preventing the learning rate from excess increase. An Artificial Neural Net'NOrk (ANN) based controller system is simulated and implemented in the laboratory using a small 48 Watts de permanent magnet motor fed from a single phase, full wave, half controlled thyristor rectifier. Turbo C ver. 2.0 compiler is used to program the controller under PC-DOS based platform. A Programmable Logic Controller (PLC) is connected to the computer using a special serial cable, which they exchange data through. The computer is talking to the PLC using its very special protocol, which is also programmed